memori

What it is and what it does

Memori is a Python SDK that intercepts LLM interactions and automatically extracts, stores, and recalls structured memory across conversations and sessions. It works by registering with supported LLM clients, then transparently capturing conversation history, tool calls, and agent decisions without requiring code changes to your existing LLM calls.

The package is designed for AI agents and multi-turn applications that need persistent context without inflating prompt sizes. It attributes memories to entities (users, systems) and processes (agents, workflows), allowing fine-grained recall and scoping. Memori can run against its cloud API (zero-config) or your own database via BYODB mode. It integrates with frameworks and supports MCP clients for developer-focused memory.

Use it for:

• Build support agents that remember customer history and preferences across sessions without manual context management
• Enable multi-step AI workflows to persist decisions and tool outputs so agents can reason over their own execution history
• Reduce LLM token costs by storing structured memory instead of repeating full conversation history in every prompt
• Implement team-wide agent memory so new engineers inherit shared context and project conventions without tribal knowledge transfer
• Connect developer tools via MCP so your coding assistant learns project conventions and coding style over time

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Memori is a Python SDK that automatically captures and recalls structured memory from LLM interactions and agent conversations, persisting context across sessions without requiring code changes to existing LLM calls.

Yes, with conditions. Memori is worth installing if you need persistent agent memory and are willing to adopt its API key model and attribution pattern. The package is actively maintained with no known vulnerabilities and offers genuine efficiency gains. The main friction is the 8 runtime dependencies and the requirement to sign up for Memori Cloud or manage your own database. If you're building stateless LLM applications or don't need cross-session memory, it adds unnecessary overhead.

Install

memori on PyPI

pip install memori

uv add memori

poetry add memori

Installing memori

Before you install

Medium install friction due to 8 runtime dependencies including aiohttp, botocore, faiss-cpu, grpcio, numpy, and protobuf. Package is actively maintained with recent releases and has 15950 GitHub stars, indicating solid community adoption. Requires Python 3.10 or later.

License in practice

Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for production deployments in most contexts.

Quickstart

pip install memori

from memori import Memori

mem = Memori().llm.register(client)
mem.attribution(entity_id="user_123", process_id="support_agent")

# Memori persists and recalls context automatically from LLM interactions

Requires MEMORI_API_KEY environment variable set; sign up at app.memorilabs.ai to obtain credentials. Also requires an LLM provider API key configured separately.

Verify before relying

• Whether faiss-cpu is required for all use cases or only for specific memory backends
• Performance overhead of automatic memory capture on LLM request latency
• Data retention and privacy guarantees for memories stored in Memori Cloud
• Which specific LLM providers are supported beyond those mentioned in documentation

Package facts

Evidence: memori-3.3.6-cp310-abi3-android_24_arm64_v8a.whl; memori-3.3.6-cp310-abi3-android_24_x86_64.whl; memori-3.3.6-cp310-abi3-macosx_11_0_arm64.whl; memori-3.3.6-cp310-abi3-macosx_11_0_x86_64.whl; memori-3.3.6-cp310-abi3-manylinux_2_28_aarch64.whl; memori-3.3.6-cp310-abi3-manylinux_2_28_x86_64.whl; memori-3.3.6-cp310-abi3-musllinux_1_2_aarch64.whl; memori-3.3.6-cp310-abi3-musllinux_1_2_x86_64.whl; memori-3.3.6-cp310-abi3-win_amd64.whl

Keywords: ai, memory, agents, llm, artificial-intelligence, multi-agent

Tags

More Python Modules packages

Further reading